3
$\begingroup$

I have a confusion matrix with 7 classes and would like to represent the matrix in a graph. Something similar like a confusion wheel.

Mainly I need to show, the correct observations in each class and incorrect ones with other classes in a pie chart.

Any python implementations would be great.

confusion wheel http://www.cvast.tuwien.ac.at/ConfusionAnalysis/

$\endgroup$
3
$\begingroup$

I would strongly recommend against using a confusion wheel to visualize your confusion matrix. As impressive and fancy as they look, confusion wheels are visually complicated and unintuitive to read. Good data visualizations summarize information in a way that is simple, clear, and intuitive. Confusion wheels have none of those properties.

Unfortunately, there really isn't a great way to visualize a large confusion matrix. The best ones I've seen are a grid-based plot with either color-coding or size-coding representing the value in each table cell. Funnily enough, the visualization in Fig. 1(a) of the link you included is significantly easier to read and more informative than the confusion wheel in Fig 1(b). A grid graph like the one in the link's Fig. 1(a) is what I would recommend.

| improve this answer | |
$\endgroup$
  • $\begingroup$ I don't think a 7 by 7 matrix is a large one. Yes, I agree that the grid-based one with variations in intensity is a good one. But that is just a replica of my confusion matrix. I need a pie chart like representation with each part showing the different areas of observations with color. $\endgroup$ – yazhi Mar 23 '16 at 16:09
  • $\begingroup$ I'm not sure why you are so focused on having a pie chart representation. Pie charts are almost never a good choice for visualization. What are you actually trying to communicate with your graph? $\endgroup$ – Data Science Dojo Mar 23 '16 at 22:27

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.